Baidu is looking to build what it describes as the world’s largest autonomous ride-hailing service area in 2023. With it, Baidu is hoping to expand the operation area for its driverless robotaxis while developing its consumer base.

The announcement was made at Apollo Day, Baidu’s autonomous driving technology event, where it unveiled new products and revealed updates for its existing product line-up. The event’s new products included an AI big model for autonomous driving perception, new HD maps, a closed-loop autonomous driving data system, and the completion of end-to-end adaptation of AI chips for autonomous vehicles.

Baidu’s AI big model for autonomous driving provides a pre-trained visual-language model that is expected to require minimal human supervision. With support from Baidu’s WenXin Big Model – which can already recognize many objects – it helps enlarge the scope of semantic recognition. The model will ultimately enable Baidu’s AVs to quickly recognize unseen objects, such as emergency service vehicles and plastic bags, among others. The company’s autonomous driving perception model, a sub-model of the WenXin Big Model, utilizes more than a billion parameters to improve the generalization potential of autonomous driving perception.

The company’s new HD maps for autonomous driving offer a 96% automation rate for map production, helping increase efficiency while reducing costs. Its ability to offer real-time map updates through integrating vehicle-side perception data and multi-source maps helps reaffirm safety. This, in addition to the data already accumulated on Baidu Maps, alongside millions of human-drive hours, works to enhance the reliability of Baidu’s autonomous driving functions.

Apollo Loop, Baidu’s closed-loop data system, will help it identify and utilize data essential to autonomous driving. To purify the data, it leverages on-board small AI models and the cloud-based big model to achieve high-efficiency data mining and automated labeling. The system’s data ingestion architecture achieves automated training with its group-optimization ability, and data distribution understanding, to utilize data effectively and further enhance the overall intelligence of its autonomous driving functions.

In mapping its future with these products, an autonomous expert from Baidu detailed how the company envisions a beneficial relationship between them. Here, L4 autonomy will continue to provide advanced technology migration for L2+ smart driving products in urban use cases, while L2 data feedback will in turn improve L4’s generalization abilities.